[Visinfo] Reading week #1 write up

[Corina S. Schweller]

Corina Schweller
MAT259
Reading Week #1

VISUALIZING KNOWLEDGE DOMAINS ? Boerner, Chen, Boyack.

The field of Domain Visualization can be very disconnected when viewed 
from different disciplines. There is a gap between theory and practice, 
which needs to be bridged. The history of databases, which are often 
employed for mapping, began in the 1950?s with citation index 
databases. In the 1960?s mapping was done manually and one of the 
pioneers was a spatial map of research in DNA. This map allows for 
scientific communication and analysis of domains. Advances of 
scientific knowledge can be shown with longitudinal mapping. This type 
of mapping can even forecast trends. A citation network can be 
navigated by SCI-Map software, which grows the map based on keywords 
and is based on clustering. Scientific Visualization is still not very 
interactive. On the other hand, Information Visualization focuses on 
interactivity. In the field of geography information can be visualized 
with geographic coordinates. In order to map information, the 
corresponding data is necessary. Then the units of analysis need to be 
selected. The most common units are documents. The Vector Space Model 
was designed for the retrieval of information. It is utilized for 
indexing documents and is composed of three parts; document indexing, 
term weighing, and computing similarity coefficients. The Vector Space 
Model works according to word matching and allows for a way to find 
similarities in documents. High dimensional data can be reduced, while 
still preserving the structure, with techniques such as the 
Eigenvalue/Eigenvector decomposition. To reduce the number of variables 
and detect relations of variables the Factor Analysis technique can be 
employed. The structure between objects in a set of proximity measure 
can be found with Multidimensional Scaling. Self-Organizing Maps 
produce a 2D map of the output layer that will show the relationship to 
the input layer. The Kohonen SOM map algorithm can organize large 
quantities of information and is used to map the Internet. Information 
can be organized in various ways. Triangulation maps random points at 
the origin of a coordinate system . Force Directed Placement sorts 
randomly placed objects and computes forces between nodes. Semantic 
Treemaps apply FDP and organize documents via clustering. Visualization 
can be outlined by the Shneiderman framework; Data Types, Typology of 
Tasks, Visualizations, and Necessary Features. Fractal Views can 
visualize large hierarchies and control the amount of information 
displayed. Less important info is removed and the number of displayed 
nodes is controlled by fractal dimension. In the future, more robust 
algorithms are needed to advance information science. More accurate 
results and a faster response will be the goal of future domain maps. I 
think mapping has brought much to a visual society and allows us to 
view data in a more comprehensible manner. The Vector Space Model seems 
like a clear method of organizing data and retrieval. With these models 
we can see  information displayed according to a method. Mapping shows 
more than just simple words, it allows us to perceive the similarities 
and differences between terms with visual spacing and connectivity.